Optimized Implementation of Memristor-Based Full Adder by Material Implication Logic Mehri Teimoory*, Amirali Amirsoleimani‡, Jafar Shamsi**, Arash Ahmadi†, Shahpour Alirezaee‡, Majid Ahmadi‡ Email: {mehri_teimoory}@yahoo.com {amirsol,alirezae,ahmadi}@uwindsor.ca {jafarshamsi}@elec.iust.ac.ir {ahmadi}@razi.ac.ir *Department of Electrical Engineering, Islamic Azad University, Science and Research Branch, Kermanshah, Iran ‡Department of Electrical and Computer Engineering, University of Windsor, Windsor, Ontario, Canada **Department of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran †Department of Electrical Engineering, Razi University, Kermanshah, Iran Abstract— Recently memristor-based applications and circuits This logic family executes basic Boolean operations by are receiving an increased attention. Furthermore, memristors applying memristor based IMPLY logic gate. are also applied in logic circuit design. Material implication logic In this paper a serial eight-bit memristor based full adder is is one of the main areas with memristors. In this paper an implemented based on material implication (IMPLY) logic. optimized memristor-based full adder design by material This full adder is designed with 27 memristors and it needs implication logic is presented. This design needs 27 memristors and less area in comparison with typical CMOS-based 8-bit full 184 computational steps for completing its operation which is adders. Also the presented full adder needs only 184 faster than previous designs [6-7]. computational steps which enhance former full adder design This paper is organized as follows. Section II introduces speed by 20 percent. IMPLY logic gate and its utilization in building different logic gates. Design and serial implementation of one-bit memristor Index Terms— Memristor, Material Implication, Full Adder, CMOS, Artificial Intelligence. based full adder as well as the proposed 8-bit adder are presented in section III. Consequently the section IV includes I. INTRODUCTION the conclusion of the paper. ogic computing has recently drawn wide attention due to II. MATERIAL IMPLICATION (IMPLY) LOGIC L emergence of memory-resistor or memristor devices [1]. Memristor’s non-volatility good scalability, low power p IMPLY q is a logic function that is called material consumption and compatibility with CMOS structures makes implication. This logic statement means “p implies q” or “if p them an ideal device for various applications. then q”. In Table 1 a truth table of material implication logic Logic operations with memristors are promising for nano is presented. This logic is similar to answering a question. p is computing circuits. Several approaches are presented for considered as a question and q as its answer. When p is wrong implementing logic circuits with memristors. Hybrid CMOS- any answer results in a true output ( logic 1) whether it is memristor based logic and stand-alone memristor based logic wrong or correct one. In other cases when p is true the answer are two main logic family [1]. Hybrid CMOS-Memristor truthfulness would determine output. Only case three of based logic is based on integrating CMOS with memristors IMPLY logic truth table as it is shown in Table 1 has FALSE only for computational purpose. Memristor Ratioed Logic (logic 0) output. In this case a correct question (p = 1) has (MRL) is presented in [2]. In this family, logical states wrong answer (q = 0). Based on the output values for each represented in voltage logic gates based on MRL are not cases of IMPLY logic this function is the equivalent to, capable to store their last output value. The other drawback in p IMP q p q (~ p) ˅ q (1) MRL design is its need to the conventional CMOS inverter gate for building the NAND and NOR gates. The stand-alone TABLE. 1. TRUTH TABLE OF IMPLY FUNCTION. memristor based logics are more promising than hybrid peer. Crossbar architecture is also one of the family logic with Case p q p IMP q memristor presented in [3], where the computation takes place in crossbar structure of memristors. Since in this family of 1 0 0 1 logic the fabrication is simple and its high density make it to 2 0 1 1 be a good candidate for future computing architectures. Other 3 1 0 0 type of innovative memristor-based logic gates are developed 4 1 1 1 and presented in [4] and [5], where the resistance represents logical state unlike MRL. Therefore they can store output of the gates as memristance of the output memristor. Both of A. IMPLY logic gate these families can perform functions inside a crossbar array but they need a sequencer to bias memristors. Material The memristor-based IMPLY logic gate constructed by two implication which is known IMPLY is presented in [4][5]. memristor (P and Q) and one resistor (RG). The polarity of the 2 Vcond Vset P Q P Q IMPLY RG P IMP Q Figure 1: Memristor-based IMPLY logic gate. memristors is depicted in Fig. 1. Also the RG connects the memristors into the ground. Two voltage biases for IMPLY logic gate, Vcond and Vset are imposed into P and Q memristors doped side respectively for P IMP Q operation. The Vcond Figure 2: The simulations of IMPLY gate for case 1. Linear ion drift model is voltage magnitude is lower than Vset (|Vcond| < |Vset|). In this applied for memristors. design the proposed IMPLY gate inputs are stored as resistance of the P and Q memristors. Interestingly, output of memristors with Vclear and Vset respectively to allocate this gate is the final changed memristance value of Q desired input to memristors. memristor. For setting input of IMPLY logic gate initial Then Vcond and Vset voltage bias sources should connect to P memristances of this memristors should be set based on the and Q to start IMPLY logic operation. Since memristance of considered input values of each case showed in Table. 1. For memristor Q (RON = 1 kΩ) is negligible in comparison with RG this purpose two bias sources are required to set initial resistance (RG = 10 kΩ), in this case voltage of the common memristance of memristors to RON and ROFF. For node of memristors and resistor would be about 1 V. The implementing logic zero memristors memristance should be in voltage across memristor Q in this case is determined through, its highest value R . The memristor should be inversely OFF RON ROFF RG biased. Therefore V voltage source is applied to set VQ Vset Vset 0 (3) clear R R 2R memristors initial memristances to OFF state. On the other OFF ON G hand, for setting logic one with memristor, its memristance Therefore the voltage across memristor Q would be zero. As a result, the memristor Q state is remained unchanged as its should be changed to its lowest value RON. Vset voltage source can be used for forward bias of the memristor. It should be previous state (logic 1). Case 4 is similar to this case. Its only mentioned both of these bias voltages should be higher than difference with case 2 is setting procedure for inputs of threshold voltage of the memristors. Also pulse width of the memristors. The voltage across memristor Q is determined by, proposed voltage should be long enough for reaching the RG VQ Vset Vcond (4) memrisance of the memristor to the considered values (RON or RON 2RG ROFF). For different input combinations stated in Table 1 IMPLY logic gate produces different outputs. In case 1, the In case 3, the memristance of Q is set to ROFF. After applying memristances of P and Q should be both set to ROFF initially. Vcond and Vset voltage bias sources the logic state remain RG, RON and ROFF are assumed 10 kΩ, 1 kΩ and 100 kΩ unchanged. The voltage across the memristor Q is, respectively. For this both memristor are biased with V , R clear V G V V (5) which is considered -1 V. Then V and V biases are Q cond cond cond set RON RG connected to P and Q memristors respectively. For simulation This voltage is not enough for changing the state of memristor. Vcond and Vset are 0.5 V and 1 V respectively. By imposing Although this voltage is forced the internal state of Q to these voltages based on voltage division between RG (RG = 10 change to RON (logic 1) it cannot be changed to its minimum kΩ) and resistance of memristor Q (ROFF = 100 kΩ), voltage of value. This is called state drift. In this state final memristance the node between two memristor and RG becomes negligible. of Q becomes lower than initial input memristance (R ). The voltage across memristor Q is, OFF RG B. Implementing logic functions by IMPLY logic VQ Vset Vcond (2) ROFF 2RG By combining IMPLY logic gate and FALSE gate any logic This makes current flows from V to ground. As the charge set structure can be produced [4-5]. FALSE gate is a logic gate passes through memristor Q its memristance is decreasing which yields zero as its output for any input. This circuit is until it reaches to its minimum value R . Therefore the final ON consisting of three memristors for implementing NAND logic memristance (R ) of Q is the desired output (logic 1) of the ON based on IMPLY and FALSE functions. P and Q are inputs IMPLY gate for case 1 of the truth table. The case 1 is and S is output. The NAND function for P and Q is written by simulated for IMPLY gate with mentioned characteristic in IMPLY and FALSE logic through, Fig. 2. This case is determined the write time of the circuit. In case 2, the memristances of P and Q should be set to ROFF S Q IMP P IMP S ~ Q ˅ ((~ P) ˅ S) (6) and R respectively.